For an Autosegmental Theory of Mutation

Autosegmental frameworks have long posited the existence of morphemes that consist, in whole or in part, of floating features, tones, or moras. Morphological feature-changing (for which “mutation” is the most common traditional label) can then be treated as concatenative in nature. Recently, it has been suggested that mutation phenomena might be brought under the purview of various approaches to handling nonconcatenative morphology in OT, of which Transderivational Anti-Faithfulness (Alderete 1999) and REALIZE-MORPHEME (Kurisu 2001) are the most notable examples, and the use of representational devices like floating features eliminated. This paper argues that a theory with floating autosegments--given the new constraints proposed here--is capable of giving a better account of attested mutation phenomena than anti-faithfulness, REALIZE-MORPHEME, or other competing approaches.Portions of this paper appear in the WCCFL 24: Proceedings of the 23rd West Coast Conference on Formal Linguistics (2005) with the title, "An Autosegmental Theory of Quirky Mutations" and is available at http://www.lingref.com/cpp/wccfl/24/paper1243.pd